skills/mukul975/anthropic-cybersecurity-skills/testing-for-xml-injection-vulnerabilities/Gen Agent Trust Hub
testing-for-xml-injection-vulnerabilities
Fail
Audited by Gen Agent Trust Hub on Apr 9, 2026
Risk Level: CRITICALDATA_EXFILTRATIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [DATA_EXFILTRATION]: The skill provides XML payloads specifically designed to exfiltrate sensitive system files (e.g., /etc/passwd) to a remote server, as documented in SKILL.md and implemented in agent.py. The agent script also contains a UTF-7 encoded payload intended to bypass security filters.
- [EXTERNAL_DOWNLOADS]: The documentation demonstrates loading external Document Type Definitions (DTDs) from remote URLs (e.g., attacker-server.com) to facilitate blind XXE attacks.
- [COMMAND_EXECUTION]: The skill provides shell commands (curl) and a Python agent script that executes automated injection attacks against user-defined web endpoints.
Recommendations
- Contains 2 malicious URL(s) - DO NOT USE
Audit Metadata